1.

Introduction

Human ex vivo skin has been used as a model of in vivo epithelia for transdermal delivery,¹ for topical penetration studies,^{2, 3} and in cosmeceutical applications.⁴ In addition, ex vivo skin is often used as a source of transplantable allografts⁴ to treat burns and scalds, providing immediate barrier protection and pain relief while stimulating reepithelialization.^{5, 6} An issue in using ex vivo skin is maintaining viability and metabolic activity, as these are required to optimally simulate in vivo conditions for percutaneous penetration and to ensure the best clinical outcome for allograft recipients.^{4, 7} The most appropriate control for assessing the viability of ex vivo skin is in vivo skin, recognizing that ex vivo skin viability decreases over time as a consequence of ischemic necrosis. This is caused by the loss of blood supply after surgical excision, hypoxia, and the accumulation of toxic metabolites.⁸

Methods commonly used to assess the viability of ex vivo skin include the morphology of the tissue,⁹ metabolic activity by MTT,^{10, 11, 12} lactate dehydrogenase (LDH) activity assays,¹³ skin pH,¹⁴ and oxygen consumption.^{11, 15, 16} Several studies have used the MTT assay to measure the decreasing viability of ex vivo skin.^{7, 8, 11, 17, 18} The MTT assay, which measures tetrazolium reduction by metabolic enzymes, is a technique widely used by researchers and skin banks for estimating the metabolic viability of ex vivo skin.^{2, 18, 19} However, the measurement of metabolic viability by tetrazolium-based assays is limited, as it does not yield direct and real-time cellular information. This assay is also limited by the necessity to destroy the tissue, which restricts clinical applicability.

Several other methodologies have been used to assess skin viability including trypan blue dye exclusion, oxygen consumption, histopathological examination, and pH change.^{11, 15, 16} Oxygen consumption and pH measurements have both been used to monitor skin metabolism noninvasively. Stratum corneum pH is considered to be critical for maintaining outer barrier function at pH 4.2 to 5.6. Therefore, alterations in skin surface pH indicate a potential disruption in that barrier. Messager ¹⁵ showed that both excised and frozen skin had a much higher pH ( $>\mathrm{pH}$ 8) and that over time the pH of the fresh skin decreased ( $<\mathrm{pH}$ 7.5), while the pH of the frozen skin did not change. Oxygen consumption is a hallmark of aerobic respiration and activity relates to the activity of mitochondria. Metabolic activity is especially critical for epidermal function. This important index of skin viability is closely related to NAD(P)H lifetime imaging, also tightly coupled to metabolic and mitochondrial activity.

Overall, both pH and oxygen consumption yield different but complementary information about skin viability to NAD(P)H imaging by multiphoton tomography and fluorescence lifetime imaging (MPT-FLIM). Oxygen consumption measurements have the advantage of low cost and ease of use, but the limitations include a lack of resolution in terms of which cells or layers of the skin are the most active or impaired. On the other hand, MPT-FLIM is a relatively expensive technique in terms of equipment, technical expertise, and data processing time required to obtain useful information, but has the advantage of subcellular resolution and multiple types of data output (i.e., microenvironment changes, semiquantitative concentration, physical distribution of metabolic species, and morphological information).

Reduced nicotinamide adenine dinucleotide (NADH) and reduced nicotinamide adenine dinucleotide phosphate (NADPH) have also been used as metabolic indicators for the viability of cells and tissues.^{20, 21, 22, 23} Recently, fluorescence lifetime imaging microscopy (FLIM) analysis of intracellular NAD(P)H has been used to monitor metabolic activity in healthy and diseased tissue.^{24, 25, 26, 27} NAD(P)H refers to the summation of the molecules NADH and NADPH. The fluorescence lifetime $\left(\tau \right)$ of a molecule is the average time the molecule stays in the excited state.²⁸ FLIM can distinguish between compounds with similar fluorescence spectral properties and is concentration independent, only changing with temperature, pH, and molecular interactions.²⁹

Combined with multiphoton tomography, MPT-FLIM is a powerful means of analyzing fluorophore populations and interactions in tissues. Traditional confocal microscopy is a linear optical process that uses a single-photon to excite a fluorophore and measure the subsequent fluorescent emission. The benefits of this method are low cost, compared to MPT systems, and widespread use. Single-photon confocal microscopy suffers from some drawbacks including scattering, photobleaching, limited depth penetration, and photodamage.³⁰ Multiphoton tomography, a much more expensive imaging technology, significantly reduces the effect of photobleaching and photodamage while increasing the depth at which tissues can be imaged.^{30, 31} While cost and complexity of running MPT systems has hindered widespread use, there are significant technological advantages. The MPT-FLIM system can measure fluorescence intensity, separate components of the skin, and quantify the fluorescence lifetime of NAD(P)H without significant variation within or between samples (Fig. 1 ). Figure 1 also includes photon-normalized spectra redrawn from previous work.^{32, 33}

Fig. 1

Fluorescence intensity and NAD(P)H lifetime intra and inter sample variation $(n=6)$ in study subjects. The images were acquired with a scan speed of $512\times 512\text{pixels}∕13.6\mathrm{s}$ at $52\times 52\mu \mathrm{m}$ in size. (a) The average fluorescence intensity and standard deviation (SD) were calculated for the six subjects. (b) The photon contribution variability of NAD(P)H, keratin and FAD; the inset illustrates the photon normalized spectra redrawn from Zvyagin ³³ (c) and (d) The mean and SD of the average fast $\left({\tau }_1\right)$ and slow $\left({\tau }_2\right)$ NAD(P)H fluorescence lifetime decays, respectively.

The fluorescence lifetimes for free and protein-bound NAD(P)H differ and are approximately 400 and $2000\mathrm{ps}$ , respectively,^{26, 29, 34, 35} with the ratio of free to protein-bound NAD(P)H being used as an indicator of the cellular $\mathrm{NADH}∕\mathrm{NAD}+$ ratio state.³⁶ These data have been used to characterize NADH in breast cancer in vitro cell lines³⁶ as well as to differentiate between normal, precancerous, and cancerous epithelia of various grades^{37, 38, 39, 40} and to define metabolic changes during apoptosis and necrosis.²¹

Both NADH and NADPH share the same absorption and fluorescence spectra,⁴¹ though neither ${\mathrm{NAD}}^{+}$ nor ${\mathrm{NADP}}^{+}$ are fluorescent.^{41, 42} Additionally, these molecules possess the same fluorescence lifetime and are often used for live metabolic and free:bound NAD(P)H ratio imaging.^{26, 41} NADH is a central coenzyme molecule in several energy metabolic pathways, including anaerobic glycolysis, the electron transport chain, and the citric acid cycle.⁴³ Intracellular NADH concentration and distribution are useful and natural metabolic indicators because they are sensitive to pathological and physiological changes.^{36, 40, 44} NADPH is a reducing agent that can counter the accumulation of reactive oxygen species²⁰ (ROS).

Studies have concluded that the relative fractions of free and protein-bound NAD(P)H, which can be inferred from the measured fluorescence lifetime,³⁶ can influence the cell redox state, defined by the ratio^{36, 41} $\left[\mathrm{NAD}\left(\mathrm{P}\right)\mathrm{H}\right]∕\left[\mathrm{NAD}{\left(\mathrm{P}\right)}^{+}\right]$ . Different rates of electron transfer are likely to prevail in the free and bound states.

In this study, we used the autofluorescence and fluorescence lifetime properties of NAD(P)H to monitor the metabolic deterioration of ex vivo skin stored at different temperatures in the presence and absence of nutrient media over a $7\text{-day}$ time course. Two controls were used: images of the ex vivo skin taken at day 0 and in vivo images of the ventral forearm from three volunteers. We showed that MPT-FLIM of ex vivo skin yields characteristic and reproducible imaging data that can be used to monitor ischemic necrosis. These observations may contribute to the development of an optical assay for clinicians and researchers to assess the viability of skin flaps, preserved skin for transplants as well as ex vivo skin for experimentation that is metabolically comparable to in vivo skin.

2.

Results

2.1.

Autofluorescence Live Images

Figure 2 shows representative autofluorescence images of in vivo and ex vivo skin, at day 0, and ex vivo skin stored in the presence and absence of media at the indicated temperatures over the $7\text{-day}$ time course. Ex vivo skin stored at $4°\mathrm{C}$ in the absence of media maintained the most consistent autofluorescence with time (Fig. 2). Skin stored at $4°\mathrm{C}$ in media and at either $-20°\mathrm{C}$ or room temperature without media showed less consistency. Autofluorescence was lost rapidly under the other conditions studied.

Fig. 2

Autofluorescence of ex vivo skin kept at $4°\mathrm{C}$ , room temperature (RT), and $37°\mathrm{C}$ in the presence and absence of nutrient media. The scale bar represents a distance of $10\mu \mathrm{m}$ . The skin depths of the images range between 30 and $50\mu \mathrm{m}$ . The absence of native cellular autofluorescence is indicated by N.D. (no detection).

2.2.

Autofluorescence Intensity and Photon Counts of NAD(P)H, Keratin, and FAD

Fluorescence intensity and NAD(P)H photon counts (total and normalized to either FAD or keratin photon counts) for each of the samples are shown in Fig. 3 . Autofluorescence intensity and NAD(P)H photon counts (total and normalized) were relatively constant over time at $4°\mathrm{C}$ and at room temperature except at day 7. In contrast, tissue incubated at $37°\mathrm{C}$ exhibited a significant decrease $(p<0.05)$ in autofluorescence intensity at day 1, which progressively decreased to zero over time (Fig. 3). A similar pattern was seen for the FAD-normalized NAD(P)H photon count with temperature and over time.

Fig. 3

Autofluorescence intensity and photon counts of NAD(P)H, keratin and FAD from ex vivo skin stored at $4°\mathrm{C}$ , room temperature, and $37°\mathrm{C}$ . The mean fluorescence intensities of the autofluorescence images from $350\text{to}620\mathrm{nm}$ are shown in the first row. The second row plots the mean photon counts of NAD(P)H, as obtained from FLIM channel 1 $\left(350\text{to}450\mathrm{nm}\right)$ . The third row shows normalized NAD(P)H photon counts to keratin photon counts, determined from FLIM channel 2 $\left(450\text{to}515\mathrm{nm}\right)$ . The final row shows normalized NAD(P)H photon counts to FAD photon counts, calculated from FLIM channel 3 $\left(515\text{to}620\mathrm{nm}\right)$ . The error bars represent the SD of the means from three independent ex vivo skin samples.

2.3.

Fluorescence Lifetime Distribution Histograms

Figure 4 shows the histograms of the lifetimes of the fast [ ${\tau }_1$ , free NAD(P)H] and slow decay components [ ${\tau }_2$ , protein-bound NAD(P)H] for both the in vivo and ex vivo controls. There was a major peak at $\sim 400\mathrm{ps}$ for both in vivo and ex vivo lifetime histograms for free NAD(P)H. A secondary peak at $\sim 800\mathrm{ps}$ was seen in ex vivo but not in vivo skin. A peak was also seen at $\sim 2500\mathrm{ps}$ with an additional peak at $\sim 4500\mathrm{ps}$ . In general, ex vivo skin samples kept at $4°\mathrm{C}$ for $7\text{days}$ without media showed two histogram peaks at $\sim 500\mathrm{ps}$ and $\sim 2500\text{to}3500\mathrm{ps}$ , comparable to the in vivo and ex vivo controls. Other samples showed much less consistency and other multiple peaks.

Fig. 4

Free $\left({\tau }_1\right)$ and protein-bound $\left({\tau }_2\right)$ NAD(P)H fluorescence lifetime histograms of in vivo and ex vivo skin stored at $4°\mathrm{C}$ , room temperature, and $-20°\mathrm{C}$ in the presence and absence of media. The fluorescence lifetime histograms of free and protein-bound NAD(P)H from FLIM imaged ex vivo skin are shown from day 7 of the time course. The separate histograms in each box represent ex vivo skin samples from up to three individuals.

2.4.

Free:Bound NAD(P)H Lifetime Ratio Images

Figure 5 shows representative pseudocolored images, between the ${\tau }_m$ range of $0\text{to}2000\mathrm{ps}$ , of in vivo control and ex vivo skin samples stored over $7\text{days}$ . The in vivo and ex vivo controls had a mean ${\tau }_m$ fluorescence lifetime of $\sim 1000\mathrm{ps}$ . Ex vivo skin stored at $4°\mathrm{C}$ , without media, maintained this mean lifetime up to $7\text{days}$ . With media, mean lifetime of skin kept at $4°\mathrm{C}$ was comparable to the controls only up to $4\text{days}$ . For the remaining conditions, the mean lifetime progressively approached $\sim 2000\mathrm{ps}$ until tissue expiration.

Fig. 5

Fluorescence lifetime images of ex vivo skin kept in the presence and absence of media, at various temperatures, across a $7\text{-day}$ time course. The images have been pseudocolored according to ${\tau }_m$ , which is calculated by weighting the mean fluorescence lifetime of free and protein-bound NAD(P)H ( ${\tau }_1$ and ${\tau }_2$ , respectively) by their respective contributions: ${\alpha }_1$ and ${\alpha }_2$ . These are representative images taken from three independent experiments.

2.5.

Image Quality

Figure 6 displays the average image quality measurement (IQM) values over time of the imaged skin for each storage condition. Image quality for all the storage conditions initially increased after $24\mathrm{h}$ relative to the day 0 control. For skin stored at $4°\mathrm{C}$ , with and without media, this value fell to the control level and persisted up until day 7. Figure 6 also shows that mean peak signal-to-noise ratio (PSNR) values from the ex vivo skin images stored at different conditions did not significantly change with time. However, the mean PSNR was higher for skin stored at $37°\mathrm{C}$ , without media than at $4°\mathrm{C}$ and room temperature without media, after $4\text{days}$ . The noise frequency spectrum profiles are displayed in Fig. 7 . Figure 7 aligns the autofluorescence images adjacent to their respective noise frequency spectra. Autofluorescence images representing viable epidermal cells (i.e., the controls and ex vivo skin at $4°\mathrm{C}$ ) all displayed an L-shaped noise frequency spectrum. In contrast, tissues with high background/noise typical of cells damaged due to ischemic necrosis, such as ex vivo skin kept at $37°\mathrm{C}$ and $-20°\mathrm{C}$ , showed a flat noise frequency spectrum.

Fig. 6

Mean PSNRs of the autofluorescence images $(\pm \mathrm{SD})$ , relative to the day 0 control (upper) and “image quality” mean from ex vivo skin samples $(\pm \mathrm{SD})$ , from the three study subjects, across the $7\text{-day}$ time course (lower). PSNRs were determined by ImageJ using a custom script. Autofluorescence images (Fig. 2) were processed with the image processing software Image Quality from a Natural Scene (MITRE, Bedford, Massachusetts) to obtain a quantitative IQM value.

Fig. 7

Noise frequency spectra of in vivo skin and ex vivo skin at day 0, $4\text{days}$ at $37°\mathrm{C}$ (no media), and $7\text{days}$ at $4°\mathrm{C}$ (no media), shown next to the corresponding representative autofluorescence image. The spectra were obtained by calculating the Fourier transform of the autofluorescence images using the software package Imatest (Imatest LCC, Boulder, Colorado). Data were collected from three independent experiments.

3.

Discussion

Previous studies demonstrate that cell death can alter NAD(P)H autofluorescence intensity in human cells. For example, autofluorescence increases during the early stages of apoptosis in primary human epithelial cells treated with cisplatin.²² Ischemia has been well described to cause a physical reduction in NAD(P)H and a corresponding decrease in autofluorescence in a variety of cells and tissues.^{23, 45, 46} The correlation with NADH autofluorescence intensity is related⁴⁷ to the physical concentration of the NADH. NADH autofluorescence has been used to assess the onset of skin flap necrosis in rats and showed a significant decrease in autofluorescence with time and necrosis.⁴⁸ In this study, a progressive loss of tissue autofluorescence intensity and FAD-normalized NAD(P)H photon counts occurred for ex vivo skin kept at $37°\mathrm{C}$ and to a lesser extent at room temperature. Ex vivo skin kept at $4°\mathrm{C}$ maintained autofluorescence across the $7\text{-day}$ time course, which agrees with previous findings^{4, 11, 18, 49} based on the MTT assay that the metabolic viability of ex vivo skin was prolonged at $4°\mathrm{C}$ .

Our results also appear to parallel the decrease in NAD(P)H autofluorescence in yeast cells after reversible injury by the addition of ${\mathrm{H}}_2{\mathrm{O}}_2$ or $\mathrm{O}\mathrm{N}\mathrm{O}{\mathrm{O}}^{-}$ to stimulate oxidative stress.²³ As the epidermal cells in our experiment were progressing through ischemic necrosis, the reduction in autofluorescence may represent an initial depletion of NAD(P)H while the cells respond to metabolic demands and antioxidant responses, which are also associated with cell death.^{23, 50, 51} In the ex vivo skin we examined, the recovery of autofluorescence and thus NAD(P)H levels may be due to a temporary burst of metabolic activity stimulated by ischemic-injury-induced damage and response systems.

In contrast, hepatocytes undergo an initial increase in autofluorescence from NADH during the early stages of hypoxia following reversible hypoxic injury.⁴⁶ When hepatocytes were exposed to longer hypoxic conditions, the autofluorescence gradually decreased from a peak level below the control until irreversible damage and cell death due to ischemia had occurred.⁴⁶ These changes occurred within $60\text{to}90\min$ after hypoxia. A similar trend was seen in this study, with the autofluorescence of the ex vivo skin initially increasing after $24\text{to}48\mathrm{h}$ incubation at $4°\mathrm{C}$ and room temperature (Fig. 2). Speculatively, the increase in autofluorescence may be due to hypoxic injury or alterations in metabolic pathways. By the end of the $7\text{-day}$ time course, all autofluorescence had been lost for tissue kept at room temperature, indicating cell death. In skin, this peak occurred at $4\text{days}$ in both the room temperature and $37°\mathrm{C}$ groups (Fig. 3). Although the trend in autofluorescence is similar in hepatocytes and skin, the time scale and temperature aspects do not clearly correlate. More investigation is required to fully understand the mechanisms that contribute to this phenomenon.

The absence of oxygen delivery to the viable epidermis in ex vivo skin may cause a significant dependence on anaerobic respiration for ATP production. The by-product of increased glycolysis is the production of lactate, which has been demonstrated to stimulate the production of reactive oxygen species.⁵² Studies that have found lactate-stimulated ROS generation noted that this occurred while some mitochondrial respiration was active. A similar effect may be occurring in viable epidermis cells as the remaining oxygen levels within the skin, postexcision, would be depleted over time with further metabolic activity. Mitochondrial respiration would eventually be superseded by glycolysis in the viable epidermis as oxygen levels decrease and lactate production increases. This transition period may be a source of ROS generation and subsequent oxidative injury that takes place in the skin.

The loss of NAD(P)H autofluorescence was not seen in skin kept at $4°\mathrm{C}$ (Fig. 2). This may be due to the low temperature that significantly suppressed any metabolic activity,⁸ irrespective of the presence of media. Generally, cellular autofluorescence for skin kept at room temperature and $37°\mathrm{C}$ lasted longer in the absence of media (Fig. 2). This may be explained by the supply of metabolic nutrients in media to promote higher metabolic activity, and thus the accumulation of metabolic waste. If this were the case, the presence of media would accelerate the path toward ischemic necrosis. This hypothesis has been suggested in a similar study conducted by others.⁸

The mean lifetimes of free and protein-bound NAD(P)H ( $\sim 400$ and $2500\mathrm{ps}$ , respectively) closely matched other in vitro measurements.^{27, 29} Additionally, the values of the in vivo and ex vivo stratum spinosum fluorescence lifetimes $\left({\tau }_2\right)$ agreed with another study.⁵³ Throughout the time course we observed the emergence of an additional ${\tau }_1$ peak lifetime between 800 and $900\mathrm{ps}$ in addition to the control peak at $\sim 400\mathrm{ps}$ in some skin samples (Fig. 4). The appearance of this peak was a common pattern for skin that had lost autofluorescence prior to tissue expiration. Similarly, for protein-bound NAD(P)H, a broad right-tailed histogram peak at $\sim 2500\mathrm{ps}$ was also observed. The formation of these additional lifetime profiles may correlate with the interaction of NAD(P)H with an intracellular enzyme. This may be similar to the different lifetimes seen for NAD(P)H when bound to lactose dehydrogenase or mitochondrial malate dehydrogenase²⁵ (mMDH). Interestingly, the recorded ${\tau }_1$ fluorescence lifetime of a solution of NADH:mMDH is²⁵ $\sim 800\mathrm{ps}$ . Thus, the formation of the $800\mathrm{ps}$ ${\tau }_1$ peak in Fig. 4 may be indicative of NADH binding to mMDH. There are also suggestions^{36, 40, 44} that similar ${\tau }_1$ peaks can be attributed to extended and folded populations of free NADH. However, the relationships between these observations and the death of skin are not known. Similarly, the appearance of a $\sim 3500\text{-}\mathrm{ps}$ peak in the ${\tau }_2$ histogram may be indicative of NADPH binding to NADPH oxidase (Nox2), which displays a comparable lifetime for this interaction.⁵⁴ Interestingly, Nox2 is involved in the production of ROS in endothelial cells in response to various agonists and stimulants.⁵⁵ One study also found that Nox2 −/− mice displayed impaired neovascularization in response to ischemia, as Nox2 is involved in angiogenesis within endothelial cells via ROS production.⁵⁵ Further studies are necessary to determine whether these hypothetical interactions occur under these conditions.

The increase in the fluorescence lifetimes of NAD(P)H for skin undergoing ischemic necrosis contrasts with the decrease in lifetime recorded for in vitro cells treated with potassium cyanide (KCN), a metabolic inhibitor.³⁶ The authors concluded³⁶ that the decrease of fluorescence lifetime correlated with an increase in the $\mathrm{NADH}∕{\mathrm{NAD}}^{+}$ ratio, potentially due to the underutilization of NADH resulting from the inhibition of the electron transport chain by KCN. In this study however, the ex vivo skin suffering from ischemic necrosis does not have a fresh source of nutrients and a means of extracting waste. Thus, the lack of nutrients during continued metabolic activity may inhibit the replenishing of fresh NADH/NADPH and lead to the accumulation of ${\mathrm{NAD}}^{+}∕{\mathrm{NADP}}^{+}$ .

FLIM images of cultured cells pseudocolored by the mean fluorescence lifetime and weighted by relative amplitudes of components $\left({\tau }_m\right)$ have been used as a visual indicator of the cellular free:bound NAD(P)H ratio state and related to cancer progression.^{36, 38} In our study, the progressive increase of the mean ${\tau }_m$ lifetime from $\sim 1000\text{to}\sim 2000\mathrm{ps}$ appears to be an indicator of ischemic necrosis. Interestingly, these results contrast with other studies that have described the average ${\tau }_m$ lifetime decreasing with hypoxia in cancerous tissue.^{38, 39} While a marked decrease in the supply of oxygen during metabolic activity is common to these physiological states, the contrast in fluorescence lifetimes may be explained by the finding that human epidermal skin metabolism is already predominantly anaerobic.⁵⁶ Thus, the lifetime increases throughout the onset of ischemic necrosis may be more related to removal of metabolic waste from the tissue via the vasculature rather than the lack of oxygen supply. Further studies are required to examine this hypothesis.

Noise frequency spectra have been largely used to assess noise compensation or filter software for digital cameras, with flat spectra indicating uncompensated white noise.⁵⁷ Cameras that use digital software to compensate for white noise display a roll-off in noise frequency spectra, characteristic of spectral noise.⁵⁷ We found that tissue with advanced ischemic necrosis gave a cloudy image, presumably due to leaking intracellular material.⁵⁸ In addition, there was a flat noise frequency spectrum rather than the L-shaped spectrum seen for the control images and the best-preserved tissue stored at $4°\mathrm{C}$ up to $7\text{days}$ (Fig. 7).

In summary, this study has examined changes in NAD(P)H autofluorescence following the progressive onset of ischemic necrosis in ex vivo skin. We have found that ischemic necrosis of the tissue is marked by several characteristics: (1) a loss in NAD(P)H autofluorescence, possibly due to a physical depletion of the molecule; (2) a reduction in the FAD normalized NAD(P)H photon count; (3) a significant deviation in the fluorescence lifetime histogram of free and protein-bound NAD(P)H away from a normal “control” state (i.e., freshly excised ex vivo skin); (4) an increase in the ${\tau }_m$ average fluorescence lifetime above $\sim 1000\mathrm{ps}$ ; and (5) a qualitative loss is cellular integrity, revealed by noise frequency spectra analysis. These patterns may serve as the basis for MPT-FLIM to be used as a visual noninvasive biopsy for assessing the onset of ischemic necrosis of ex vivo skin.

4.

Materials and Methods